Today’s vocab word can mean a lot of things in different contexts, but it almost always means the center or core. A kernel can refer to the edible interior of a nut or seed. In computer science, a kernel is part of an operating system.
On Spatially Challenged, I’m usually referring to kernel in a geospatial sense. In this context, kernels are functions that are used to calculate neighborhoods. Geospatial analysis relies on the concept that things are close together are more similar to each other than things that are far apart (this is Tobler’s first law of geography, and I should really do a post about it in more detail). Things that are in the same neighborhood are expected to be similar.
Carefully defining a neighborhood is very important because it lets us decide how close things need to be in order to be considered a neighbor. This effectively controls the spatial scale that our study focuses on. You define a kernel using two key elements: the shape and the bandwidth. Shape is determined using a function, such as the Gaussian function or a negative absolute value function (see my vague sticky note drawing to the left if you aren’t familiar with these). The height of the function at a given distance from the center of the neighborhood determines how much weight the given point receives. The bandwidth of the kernel determines how far the function extends from the center.
Looking to learn more about kernels or other terms? I’ve been using Geospatial Analysis 6th edition lately. It’s available online here.